summaryrefslogtreecommitdiff
path: root/arch/s390/kernel/perf_cpum_sf.c
blob: 544a02e944c65156aca41c4b8edffe2882862bfb (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
// SPDX-License-Identifier: GPL-2.0
/*
 * Performance event support for the System z CPU-measurement Sampling Facility
 *
 * Copyright IBM Corp. 2013, 2018
 * Author(s): Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
 */
#define KMSG_COMPONENT	"cpum_sf"
#define pr_fmt(fmt)	KMSG_COMPONENT ": " fmt

#include <linux/kernel.h>
#include <linux/kernel_stat.h>
#include <linux/perf_event.h>
#include <linux/percpu.h>
#include <linux/pid.h>
#include <linux/notifier.h>
#include <linux/export.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/moduleparam.h>
#include <asm/cpu_mf.h>
#include <asm/irq.h>
#include <asm/debug.h>
#include <asm/timex.h>

/* Minimum number of sample-data-block-tables:
 * At least one table is required for the sampling buffer structure.
 * A single table contains up to 511 pointers to sample-data-blocks.
 */
#define CPUM_SF_MIN_SDBT	1

/* Number of sample-data-blocks per sample-data-block-table (SDBT):
 * A table contains SDB pointers (8 bytes) and one table-link entry
 * that points to the origin of the next SDBT.
 */
#define CPUM_SF_SDB_PER_TABLE	((PAGE_SIZE - 8) / 8)

/* Maximum page offset for an SDBT table-link entry:
 * If this page offset is reached, a table-link entry to the next SDBT
 * must be added.
 */
#define CPUM_SF_SDBT_TL_OFFSET	(CPUM_SF_SDB_PER_TABLE * 8)
static inline int require_table_link(const void *sdbt)
{
	return ((unsigned long) sdbt & ~PAGE_MASK) == CPUM_SF_SDBT_TL_OFFSET;
}

/* Minimum and maximum sampling buffer sizes:
 *
 * This number represents the maximum size of the sampling buffer taking
 * the number of sample-data-block-tables into account.  Note that these
 * numbers apply to the basic-sampling function only.
 * The maximum number of SDBs is increased by CPUM_SF_SDB_DIAG_FACTOR if
 * the diagnostic-sampling function is active.
 *
 * Sampling buffer size		Buffer characteristics
 * ---------------------------------------------------
 *	 64KB		    ==	  16 pages (4KB per page)
 *				   1 page  for SDB-tables
 *				  15 pages for SDBs
 *
 *  32MB		    ==	8192 pages (4KB per page)
 *				  16 pages for SDB-tables
 *				8176 pages for SDBs
 */
static unsigned long __read_mostly CPUM_SF_MIN_SDB = 15;
static unsigned long __read_mostly CPUM_SF_MAX_SDB = 8176;
static unsigned long __read_mostly CPUM_SF_SDB_DIAG_FACTOR = 1;

struct sf_buffer {
	unsigned long	 *sdbt;	    /* Sample-data-block-table origin */
	/* buffer characteristics (required for buffer increments) */
	unsigned long  num_sdb;	    /* Number of sample-data-blocks */
	unsigned long num_sdbt;	    /* Number of sample-data-block-tables */
	unsigned long	 *tail;	    /* last sample-data-block-table */
};

struct aux_buffer {
	struct sf_buffer sfb;
	unsigned long head;	   /* index of SDB of buffer head */
	unsigned long alert_mark;  /* index of SDB of alert request position */
	unsigned long empty_mark;  /* mark of SDB not marked full */
	unsigned long *sdb_index;  /* SDB address for fast lookup */
	unsigned long *sdbt_index; /* SDBT address for fast lookup */
};

struct cpu_hw_sf {
	/* CPU-measurement sampling information block */
	struct hws_qsi_info_block qsi;
	/* CPU-measurement sampling control block */
	struct hws_lsctl_request_block lsctl;
	struct sf_buffer sfb;	    /* Sampling buffer */
	unsigned int flags;	    /* Status flags */
	struct perf_event *event;   /* Scheduled perf event */
	struct perf_output_handle handle; /* AUX buffer output handle */
};
static DEFINE_PER_CPU(struct cpu_hw_sf, cpu_hw_sf);

/* Debug feature */
static debug_info_t *sfdbg;

/*
 * sf_disable() - Switch off sampling facility
 */
static int sf_disable(void)
{
	struct hws_lsctl_request_block sreq;

	memset(&sreq, 0, sizeof(sreq));
	return lsctl(&sreq);
}

/*
 * sf_buffer_available() - Check for an allocated sampling buffer
 */
static int sf_buffer_available(struct cpu_hw_sf *cpuhw)
{
	return !!cpuhw->sfb.sdbt;
}

/*
 * deallocate sampling facility buffer
 */
static void free_sampling_buffer(struct sf_buffer *sfb)
{
	unsigned long *sdbt, *curr;

	if (!sfb->sdbt)
		return;

	sdbt = sfb->sdbt;
	curr = sdbt;

	/* Free the SDBT after all SDBs are processed... */
	while (1) {
		if (!*curr || !sdbt)
			break;

		/* Process table-link entries */
		if (is_link_entry(curr)) {
			curr = get_next_sdbt(curr);
			if (sdbt)
				free_page((unsigned long) sdbt);

			/* If the origin is reached, sampling buffer is freed */
			if (curr == sfb->sdbt)
				break;
			else
				sdbt = curr;
		} else {
			/* Process SDB pointer */
			if (*curr) {
				free_page(*curr);
				curr++;
			}
		}
	}

	debug_sprintf_event(sfdbg, 5,
			    "free_sampling_buffer: freed sdbt=%p\n", sfb->sdbt);
	memset(sfb, 0, sizeof(*sfb));
}

static int alloc_sample_data_block(unsigned long *sdbt, gfp_t gfp_flags)
{
	unsigned long sdb, *trailer;

	/* Allocate and initialize sample-data-block */
	sdb = get_zeroed_page(gfp_flags);
	if (!sdb)
		return -ENOMEM;
	trailer = trailer_entry_ptr(sdb);
	*trailer = SDB_TE_ALERT_REQ_MASK;

	/* Link SDB into the sample-data-block-table */
	*sdbt = sdb;

	return 0;
}

/*
 * realloc_sampling_buffer() - extend sampler memory
 *
 * Allocates new sample-data-blocks and adds them to the specified sampling
 * buffer memory.
 *
 * Important: This modifies the sampling buffer and must be called when the
 *	      sampling facility is disabled.
 *
 * Returns zero on success, non-zero otherwise.
 */
static int realloc_sampling_buffer(struct sf_buffer *sfb,
				   unsigned long num_sdb, gfp_t gfp_flags)
{
	int i, rc;
	unsigned long *new, *tail;

	if (!sfb->sdbt || !sfb->tail)
		return -EINVAL;

	if (!is_link_entry(sfb->tail))
		return -EINVAL;

	/* Append to the existing sampling buffer, overwriting the table-link
	 * register.
	 * The tail variables always points to the "tail" (last and table-link)
	 * entry in an SDB-table.
	 */
	tail = sfb->tail;

	/* Do a sanity check whether the table-link entry points to
	 * the sampling buffer origin.
	 */
	if (sfb->sdbt != get_next_sdbt(tail)) {
		debug_sprintf_event(sfdbg, 3, "realloc_sampling_buffer: "
				    "sampling buffer is not linked: origin=%p"
				    "tail=%p\n",
				    (void *) sfb->sdbt, (void *) tail);
		return -EINVAL;
	}

	/* Allocate remaining SDBs */
	rc = 0;
	for (i = 0; i < num_sdb; i++) {
		/* Allocate a new SDB-table if it is full. */
		if (require_table_link(tail)) {
			new = (unsigned long *) get_zeroed_page(gfp_flags);
			if (!new) {
				rc = -ENOMEM;
				break;
			}
			sfb->num_sdbt++;
			/* Link current page to tail of chain */
			*tail = (unsigned long)(void *) new + 1;
			tail = new;
		}

		/* Allocate a new sample-data-block.
		 * If there is not enough memory, stop the realloc process
		 * and simply use what was allocated.  If this is a temporary
		 * issue, a new realloc call (if required) might succeed.
		 */
		rc = alloc_sample_data_block(tail, gfp_flags);
		if (rc)
			break;
		sfb->num_sdb++;
		tail++;
	}

	/* Link sampling buffer to its origin */
	*tail = (unsigned long) sfb->sdbt + 1;
	sfb->tail = tail;

	debug_sprintf_event(sfdbg, 4, "realloc_sampling_buffer: new buffer"
			    " settings: sdbt=%lu sdb=%lu\n",
			    sfb->num_sdbt, sfb->num_sdb);
	return rc;
}

/*
 * allocate_sampling_buffer() - allocate sampler memory
 *
 * Allocates and initializes a sampling buffer structure using the
 * specified number of sample-data-blocks (SDB).  For each allocation,
 * a 4K page is used.  The number of sample-data-block-tables (SDBT)
 * are calculated from SDBs.
 * Also set the ALERT_REQ mask in each SDBs trailer.
 *
 * Returns zero on success, non-zero otherwise.
 */
static int alloc_sampling_buffer(struct sf_buffer *sfb, unsigned long num_sdb)
{
	int rc;

	if (sfb->sdbt)
		return -EINVAL;

	/* Allocate the sample-data-block-table origin */
	sfb->sdbt = (unsigned long *) get_zeroed_page(GFP_KERNEL);
	if (!sfb->sdbt)
		return -ENOMEM;
	sfb->num_sdb = 0;
	sfb->num_sdbt = 1;

	/* Link the table origin to point to itself to prepare for
	 * realloc_sampling_buffer() invocation.
	 */
	sfb->tail = sfb->sdbt;
	*sfb->tail = (unsigned long)(void *) sfb->sdbt + 1;

	/* Allocate requested number of sample-data-blocks */
	rc = realloc_sampling_buffer(sfb, num_sdb, GFP_KERNEL);
	if (rc) {
		free_sampling_buffer(sfb);
		debug_sprintf_event(sfdbg, 4, "alloc_sampling_buffer: "
			"realloc_sampling_buffer failed with rc=%i\n", rc);
	} else
		debug_sprintf_event(sfdbg, 4,
			"alloc_sampling_buffer: tear=%p dear=%p\n",
			sfb->sdbt, (void *) *sfb->sdbt);
	return rc;
}

static void sfb_set_limits(unsigned long min, unsigned long max)
{
	struct hws_qsi_info_block si;

	CPUM_SF_MIN_SDB = min;
	CPUM_SF_MAX_SDB = max;

	memset(&si, 0, sizeof(si));
	if (!qsi(&si))
		CPUM_SF_SDB_DIAG_FACTOR = DIV_ROUND_UP(si.dsdes, si.bsdes);
}

static unsigned long sfb_max_limit(struct hw_perf_event *hwc)
{
	return SAMPL_DIAG_MODE(hwc) ? CPUM_SF_MAX_SDB * CPUM_SF_SDB_DIAG_FACTOR
				    : CPUM_SF_MAX_SDB;
}

static unsigned long sfb_pending_allocs(struct sf_buffer *sfb,
					struct hw_perf_event *hwc)
{
	if (!sfb->sdbt)
		return SFB_ALLOC_REG(hwc);
	if (SFB_ALLOC_REG(hwc) > sfb->num_sdb)
		return SFB_ALLOC_REG(hwc) - sfb->num_sdb;
	return 0;
}

static int sfb_has_pending_allocs(struct sf_buffer *sfb,
				   struct hw_perf_event *hwc)
{
	return sfb_pending_allocs(sfb, hwc) > 0;
}

static void sfb_account_allocs(unsigned long num, struct hw_perf_event *hwc)
{
	/* Limit the number of SDBs to not exceed the maximum */
	num = min_t(unsigned long, num, sfb_max_limit(hwc) - SFB_ALLOC_REG(hwc));
	if (num)
		SFB_ALLOC_REG(hwc) += num;
}

static void sfb_init_allocs(unsigned long num, struct hw_perf_event *hwc)
{
	SFB_ALLOC_REG(hwc) = 0;
	sfb_account_allocs(num, hwc);
}

static void deallocate_buffers(struct cpu_hw_sf *cpuhw)
{
	if (cpuhw->sfb.sdbt)
		free_sampling_buffer(&cpuhw->sfb);
}

static int allocate_buffers(struct cpu_hw_sf *cpuhw, struct hw_perf_event *hwc)
{
	unsigned long n_sdb, freq, factor;
	size_t sample_size;

	/* Calculate sampling buffers using 4K pages
	 *
	 *    1. Determine the sample data size which depends on the used
	 *	 sampling functions, for example, basic-sampling or
	 *	 basic-sampling with diagnostic-sampling.
	 *
	 *    2. Use the sampling frequency as input.  The sampling buffer is
	 *	 designed for almost one second.  This can be adjusted through
	 *	 the "factor" variable.
	 *	 In any case, alloc_sampling_buffer() sets the Alert Request
	 *	 Control indicator to trigger a measurement-alert to harvest
	 *	 sample-data-blocks (sdb).
	 *
	 *    3. Compute the number of sample-data-blocks and ensure a minimum
	 *	 of CPUM_SF_MIN_SDB.  Also ensure the upper limit does not
	 *	 exceed a "calculated" maximum.  The symbolic maximum is
	 *	 designed for basic-sampling only and needs to be increased if
	 *	 diagnostic-sampling is active.
	 *	 See also the remarks for these symbolic constants.
	 *
	 *    4. Compute the number of sample-data-block-tables (SDBT) and
	 *	 ensure a minimum of CPUM_SF_MIN_SDBT (one table can manage up
	 *	 to 511 SDBs).
	 */
	sample_size = sizeof(struct hws_basic_entry);
	freq = sample_rate_to_freq(&cpuhw->qsi, SAMPL_RATE(hwc));
	factor = 1;
	n_sdb = DIV_ROUND_UP(freq, factor * ((PAGE_SIZE-64) / sample_size));
	if (n_sdb < CPUM_SF_MIN_SDB)
		n_sdb = CPUM_SF_MIN_SDB;

	/* If there is already a sampling buffer allocated, it is very likely
	 * that the sampling facility is enabled too.  If the event to be
	 * initialized requires a greater sampling buffer, the allocation must
	 * be postponed.  Changing the sampling buffer requires the sampling
	 * facility to be in the disabled state.  So, account the number of
	 * required SDBs and let cpumsf_pmu_enable() resize the buffer just
	 * before the event is started.
	 */
	sfb_init_allocs(n_sdb, hwc);
	if (sf_buffer_available(cpuhw))
		return 0;

	debug_sprintf_event(sfdbg, 3,
			    "allocate_buffers: rate=%lu f=%lu sdb=%lu/%lu"
			    " sample_size=%lu cpuhw=%p\n",
			    SAMPL_RATE(hwc), freq, n_sdb, sfb_max_limit(hwc),
			    sample_size, cpuhw);

	return alloc_sampling_buffer(&cpuhw->sfb,
				     sfb_pending_allocs(&cpuhw->sfb, hwc));
}

static unsigned long min_percent(unsigned int percent, unsigned long base,
				 unsigned long min)
{
	return min_t(unsigned long, min, DIV_ROUND_UP(percent * base, 100));
}

static unsigned long compute_sfb_extent(unsigned long ratio, unsigned long base)
{
	/* Use a percentage-based approach to extend the sampling facility
	 * buffer.  Accept up to 5% sample data loss.
	 * Vary the extents between 1% to 5% of the current number of
	 * sample-data-blocks.
	 */
	if (ratio <= 5)
		return 0;
	if (ratio <= 25)
		return min_percent(1, base, 1);
	if (ratio <= 50)
		return min_percent(1, base, 1);
	if (ratio <= 75)
		return min_percent(2, base, 2);
	if (ratio <= 100)
		return min_percent(3, base, 3);
	if (ratio <= 250)
		return min_percent(4, base, 4);

	return min_percent(5, base, 8);
}

static void sfb_account_overflows(struct cpu_hw_sf *cpuhw,
				  struct hw_perf_event *hwc)
{
	unsigned long ratio, num;

	if (!OVERFLOW_REG(hwc))
		return;

	/* The sample_overflow contains the average number of sample data
	 * that has been lost because sample-data-blocks were full.
	 *
	 * Calculate the total number of sample data entries that has been
	 * discarded.  Then calculate the ratio of lost samples to total samples
	 * per second in percent.
	 */
	ratio = DIV_ROUND_UP(100 * OVERFLOW_REG(hwc) * cpuhw->sfb.num_sdb,
			     sample_rate_to_freq(&cpuhw->qsi, SAMPL_RATE(hwc)));

	/* Compute number of sample-data-blocks */
	num = compute_sfb_extent(ratio, cpuhw->sfb.num_sdb);
	if (num)
		sfb_account_allocs(num, hwc);

	debug_sprintf_event(sfdbg, 5, "sfb: overflow: overflow=%llu ratio=%lu"
			    " num=%lu\n", OVERFLOW_REG(hwc), ratio, num);
	OVERFLOW_REG(hwc) = 0;
}

/* extend_sampling_buffer() - Extend sampling buffer
 * @sfb:	Sampling buffer structure (for local CPU)
 * @hwc:	Perf event hardware structure
 *
 * Use this function to extend the sampling buffer based on the overflow counter
 * and postponed allocation extents stored in the specified Perf event hardware.
 *
 * Important: This function disables the sampling facility in order to safely
 *	      change the sampling buffer structure.  Do not call this function
 *	      when the PMU is active.
 */
static void extend_sampling_buffer(struct sf_buffer *sfb,
				   struct hw_perf_event *hwc)
{
	unsigned long num, num_old;
	int rc;

	num = sfb_pending_allocs(sfb, hwc);
	if (!num)
		return;
	num_old = sfb->num_sdb;

	/* Disable the sampling facility to reset any states and also
	 * clear pending measurement alerts.
	 */
	sf_disable();

	/* Extend the sampling buffer.
	 * This memory allocation typically happens in an atomic context when
	 * called by perf.  Because this is a reallocation, it is fine if the
	 * new SDB-request cannot be satisfied immediately.
	 */
	rc = realloc_sampling_buffer(sfb, num, GFP_ATOMIC);
	if (rc)
		debug_sprintf_event(sfdbg, 5, "sfb: extend: realloc "
				    "failed with rc=%i\n", rc);

	if (sfb_has_pending_allocs(sfb, hwc))
		debug_sprintf_event(sfdbg, 5, "sfb: extend: "
				    "req=%lu alloc=%lu remaining=%lu\n",
				    num, sfb->num_sdb - num_old,
				    sfb_pending_allocs(sfb, hwc));
}

/* Number of perf events counting hardware events */
static atomic_t num_events;
/* Used to avoid races in calling reserve/release_cpumf_hardware */
static DEFINE_MUTEX(pmc_reserve_mutex);

#define PMC_INIT      0
#define PMC_RELEASE   1
#define PMC_FAILURE   2
static void setup_pmc_cpu(void *flags)
{
	int err;
	struct cpu_hw_sf *cpusf = this_cpu_ptr(&cpu_hw_sf);

	err = 0;
	switch (*((int *) flags)) {
	case PMC_INIT:
		memset(cpusf, 0, sizeof(*cpusf));
		err = qsi(&cpusf->qsi);
		if (err)
			break;
		cpusf->flags |= PMU_F_RESERVED;
		err = sf_disable();
		if (err)
			pr_err("Switching off the sampling facility failed "
			       "with rc=%i\n", err);
		debug_sprintf_event(sfdbg, 5,
				    "setup_pmc_cpu: initialized: cpuhw=%p\n", cpusf);
		break;
	case PMC_RELEASE:
		cpusf->flags &= ~PMU_F_RESERVED;
		err = sf_disable();
		if (err) {
			pr_err("Switching off the sampling facility failed "
			       "with rc=%i\n", err);
		} else
			deallocate_buffers(cpusf);
		debug_sprintf_event(sfdbg, 5,
				    "setup_pmc_cpu: released: cpuhw=%p\n", cpusf);
		break;
	}
	if (err)
		*((int *) flags) |= PMC_FAILURE;
}

static void release_pmc_hardware(void)
{
	int flags = PMC_RELEASE;

	irq_subclass_unregister(IRQ_SUBCLASS_MEASUREMENT_ALERT);
	on_each_cpu(setup_pmc_cpu, &flags, 1);
}

static int reserve_pmc_hardware(void)
{
	int flags = PMC_INIT;

	on_each_cpu(setup_pmc_cpu, &flags, 1);
	if (flags & PMC_FAILURE) {
		release_pmc_hardware();
		return -ENODEV;
	}
	irq_subclass_register(IRQ_SUBCLASS_MEASUREMENT_ALERT);

	return 0;
}

static void hw_perf_event_destroy(struct perf_event *event)
{
	/* Release PMC if this is the last perf event */
	if (!atomic_add_unless(&num_events, -1, 1)) {
		mutex_lock(&pmc_reserve_mutex);
		if (atomic_dec_return(&num_events) == 0)
			release_pmc_hardware();
		mutex_unlock(&pmc_reserve_mutex);
	}
}

static void hw_init_period(struct hw_perf_event *hwc, u64 period)
{
	hwc->sample_period = period;
	hwc->last_period = hwc->sample_period;
	local64_set(&hwc->period_left, hwc->sample_period);
}

static void hw_reset_registers(struct hw_perf_event *hwc,
			       unsigned long *sdbt_origin)
{
	/* (Re)set to first sample-data-block-table */
	TEAR_REG(hwc) = (unsigned long) sdbt_origin;
}

static unsigned long hw_limit_rate(const struct hws_qsi_info_block *si,
				   unsigned long rate)
{
	return clamp_t(unsigned long, rate,
		       si->min_sampl_rate, si->max_sampl_rate);
}

static u32 cpumsf_pid_type(struct perf_event *event,
			   u32 pid, enum pid_type type)
{
	struct task_struct *tsk;

	/* Idle process */
	if (!pid)
		goto out;

	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
	pid = -1;
	if (tsk) {
		/*
		 * Only top level events contain the pid namespace in which
		 * they are created.
		 */
		if (event->parent)
			event = event->parent;
		pid = __task_pid_nr_ns(tsk, type, event->ns);
		/*
		 * See also 1d953111b648
		 * "perf/core: Don't report zero PIDs for exiting tasks".
		 */
		if (!pid && !pid_alive(tsk))
			pid = -1;
	}
out:
	return pid;
}

static void cpumsf_output_event_pid(struct perf_event *event,
				    struct perf_sample_data *data,
				    struct pt_regs *regs)
{
	u32 pid;
	struct perf_event_header header;
	struct perf_output_handle handle;

	/*
	 * Obtain the PID from the basic-sampling data entry and
	 * correct the data->tid_entry.pid value.
	 */
	pid = data->tid_entry.pid;

	/* Protect callchain buffers, tasks */
	rcu_read_lock();

	perf_prepare_sample(&header, data, event, regs);
	if (perf_output_begin(&handle, event, header.size))
		goto out;

	/* Update the process ID (see also kernel/events/core.c) */
	data->tid_entry.pid = cpumsf_pid_type(event, pid, PIDTYPE_TGID);
	data->tid_entry.tid = cpumsf_pid_type(event, pid, PIDTYPE_PID);

	perf_output_sample(&handle, &header, data, event);
	perf_output_end(&handle);
out:
	rcu_read_unlock();
}

static unsigned long getrate(bool freq, unsigned long sample,
			     struct hws_qsi_info_block *si)
{
	unsigned long rate;

	if (freq) {
		rate = freq_to_sample_rate(si, sample);
		rate = hw_limit_rate(si, rate);
	} else {
		/* The min/max sampling rates specifies the valid range
		 * of sample periods.  If the specified sample period is
		 * out of range, limit the period to the range boundary.
		 */
		rate = hw_limit_rate(si, sample);

		/* The perf core maintains a maximum sample rate that is
		 * configurable through the sysctl interface.  Ensure the
		 * sampling rate does not exceed this value.  This also helps
		 * to avoid throttling when pushing samples with
		 * perf_event_overflow().
		 */
		if (sample_rate_to_freq(si, rate) >
		    sysctl_perf_event_sample_rate) {
			debug_sprintf_event(sfdbg, 1,
					    "Sampling rate exceeds maximum "
					    "perf sample rate\n");
			rate = 0;
		}
	}
	return rate;
}

/* The sampling information (si) contains information about the
 * min/max sampling intervals and the CPU speed.  So calculate the
 * correct sampling interval and avoid the whole period adjust
 * feedback loop.
 *
 * Since the CPU Measurement sampling facility can not handle frequency
 * calculate the sampling interval when frequency is specified using
 * this formula:
 *	interval := cpu_speed * 1000000 / sample_freq
 *
 * Returns errno on bad input and zero on success with parameter interval
 * set to the correct sampling rate.
 *
 * Note: This function turns off freq bit to avoid calling function
 * perf_adjust_period(). This causes frequency adjustment in the common
 * code part which causes tremendous variations in the counter values.
 */
static int __hw_perf_event_init_rate(struct perf_event *event,
				     struct hws_qsi_info_block *si)
{
	struct perf_event_attr *attr = &event->attr;
	struct hw_perf_event *hwc = &event->hw;
	unsigned long rate;

	if (attr->freq) {
		if (!attr->sample_freq)
			return -EINVAL;
		rate = getrate(attr->freq, attr->sample_freq, si);
		attr->freq = 0;		/* Don't call  perf_adjust_period() */
		SAMPL_FLAGS(hwc) |= PERF_CPUM_SF_FREQ_MODE;
	} else {
		rate = getrate(attr->freq, attr->sample_period, si);
		if (!rate)
			return -EINVAL;
	}
	attr->sample_period = rate;
	SAMPL_RATE(hwc) = rate;
	hw_init_period(hwc, SAMPL_RATE(hwc));
	debug_sprintf_event(sfdbg, 4, "__hw_perf_event_init_rate:"
			    "cpu:%d period:%llx freq:%d,%#lx\n", event->cpu,
			    event->attr.sample_period, event->attr.freq,
			    SAMPLE_FREQ_MODE(hwc));
	return 0;
}

static int __hw_perf_event_init(struct perf_event *event)
{
	struct cpu_hw_sf *cpuhw;
	struct hws_qsi_info_block si;
	struct perf_event_attr *attr = &event->attr;
	struct hw_perf_event *hwc = &event->hw;
	int cpu, err;

	/* Reserve CPU-measurement sampling facility */
	err = 0;
	if (!atomic_inc_not_zero(&num_events)) {
		mutex_lock(&pmc_reserve_mutex);
		if (atomic_read(&num_events) == 0 && reserve_pmc_hardware())
			err = -EBUSY;
		else
			atomic_inc(&num_events);
		mutex_unlock(&pmc_reserve_mutex);
	}
	event->destroy = hw_perf_event_destroy;

	if (err)
		goto out;

	/* Access per-CPU sampling information (query sampling info) */
	/*
	 * The event->cpu value can be -1 to count on every CPU, for example,
	 * when attaching to a task.  If this is specified, use the query
	 * sampling info from the current CPU, otherwise use event->cpu to
	 * retrieve the per-CPU information.
	 * Later, cpuhw indicates whether to allocate sampling buffers for a
	 * particular CPU (cpuhw!=NULL) or each online CPU (cpuw==NULL).
	 */
	memset(&si, 0, sizeof(si));
	cpuhw = NULL;
	if (event->cpu == -1)
		qsi(&si);
	else {
		/* Event is pinned to a particular CPU, retrieve the per-CPU
		 * sampling structure for accessing the CPU-specific QSI.
		 */
		cpuhw = &per_cpu(cpu_hw_sf, event->cpu);
		si = cpuhw->qsi;
	}

	/* Check sampling facility authorization and, if not authorized,
	 * fall back to other PMUs.  It is safe to check any CPU because
	 * the authorization is identical for all configured CPUs.
	 */
	if (!si.as) {
		err = -ENOENT;
		goto out;
	}

	/* Always enable basic sampling */
	SAMPL_FLAGS(hwc) = PERF_CPUM_SF_BASIC_MODE;

	/* Check if diagnostic sampling is requested.  Deny if the required
	 * sampling authorization is missing.
	 */
	if (attr->config == PERF_EVENT_CPUM_SF_DIAG) {
		if (!si.ad) {
			err = -EPERM;
			goto out;
		}
		SAMPL_FLAGS(hwc) |= PERF_CPUM_SF_DIAG_MODE;
	}

	/* Check and set other sampling flags */
	if (attr->config1 & PERF_CPUM_SF_FULL_BLOCKS)
		SAMPL_FLAGS(hwc) |= PERF_CPUM_SF_FULL_BLOCKS;

	err =  __hw_perf_event_init_rate(event, &si);
	if (err)
		goto out;

	/* Initialize sample data overflow accounting */
	hwc->extra_reg.reg = REG_OVERFLOW;
	OVERFLOW_REG(hwc) = 0;

	/* Use AUX buffer. No need to allocate it by ourself */
	if (attr->config == PERF_EVENT_CPUM_SF_DIAG)
		return 0;

	/* Allocate the per-CPU sampling buffer using the CPU information
	 * from the event.  If the event is not pinned to a particular
	 * CPU (event->cpu == -1; or cpuhw == NULL), allocate sampling
	 * buffers for each online CPU.
	 */
	if (cpuhw)
		/* Event is pinned to a particular CPU */
		err = allocate_buffers(cpuhw, hwc);
	else {
		/* Event is not pinned, allocate sampling buffer on
		 * each online CPU
		 */
		for_each_online_cpu(cpu) {
			cpuhw = &per_cpu(cpu_hw_sf, cpu);
			err = allocate_buffers(cpuhw, hwc);
			if (err)
				break;
		}
	}

	/* If PID/TID sampling is active, replace the default overflow
	 * handler to extract and resolve the PIDs from the basic-sampling
	 * data entries.
	 */
	if (event->attr.sample_type & PERF_SAMPLE_TID)
		if (is_default_overflow_handler(event))
			event->overflow_handler = cpumsf_output_event_pid;
out:
	return err;
}

static int cpumsf_pmu_event_init(struct perf_event *event)
{
	int err;

	/* No support for taken branch sampling */
	if (has_branch_stack(event))
		return -EOPNOTSUPP;

	switch (event->attr.type) {
	case PERF_TYPE_RAW:
		if ((event->attr.config != PERF_EVENT_CPUM_SF) &&
		    (event->attr.config != PERF_EVENT_CPUM_SF_DIAG))
			return -ENOENT;
		break;
	case PERF_TYPE_HARDWARE:
		/* Support sampling of CPU cycles in addition to the
		 * counter facility.  However, the counter facility
		 * is more precise and, hence, restrict this PMU to
		 * sampling events only.
		 */
		if (event->attr.config != PERF_COUNT_HW_CPU_CYCLES)
			return -ENOENT;
		if (!is_sampling_event(event))
			return -ENOENT;
		break;
	default:
		return -ENOENT;
	}

	/* Check online status of the CPU to which the event is pinned */
	if (event->cpu >= 0 && !cpu_online(event->cpu))
			return -ENODEV;

	/* Force reset of idle/hv excludes regardless of what the
	 * user requested.
	 */
	if (event->attr.exclude_hv)
		event->attr.exclude_hv = 0;
	if (event->attr.exclude_idle)
		event->attr.exclude_idle = 0;

	err = __hw_perf_event_init(event);
	if (unlikely(err))
		if (event->destroy)
			event->destroy(event);
	return err;
}

static void cpumsf_pmu_enable(struct pmu *pmu)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);
	struct hw_perf_event *hwc;
	int err;

	if (cpuhw->flags & PMU_F_ENABLED)
		return;

	if (cpuhw->flags & PMU_F_ERR_MASK)
		return;

	/* Check whether to extent the sampling buffer.
	 *
	 * Two conditions trigger an increase of the sampling buffer for a
	 * perf event:
	 *    1. Postponed buffer allocations from the event initialization.
	 *    2. Sampling overflows that contribute to pending allocations.
	 *
	 * Note that the extend_sampling_buffer() function disables the sampling
	 * facility, but it can be fully re-enabled using sampling controls that
	 * have been saved in cpumsf_pmu_disable().
	 */
	if (cpuhw->event) {
		hwc = &cpuhw->event->hw;
		if (!(SAMPL_DIAG_MODE(hwc))) {
			/*
			 * Account number of overflow-designated
			 * buffer extents
			 */
			sfb_account_overflows(cpuhw, hwc);
			if (sfb_has_pending_allocs(&cpuhw->sfb, hwc))
				extend_sampling_buffer(&cpuhw->sfb, hwc);
		}
		/* Rate may be adjusted with ioctl() */
		cpuhw->lsctl.interval = SAMPL_RATE(&cpuhw->event->hw);
	}

	/* (Re)enable the PMU and sampling facility */
	cpuhw->flags |= PMU_F_ENABLED;
	barrier();

	err = lsctl(&cpuhw->lsctl);
	if (err) {
		cpuhw->flags &= ~PMU_F_ENABLED;
		pr_err("Loading sampling controls failed: op=%i err=%i\n",
			1, err);
		return;
	}

	/* Load current program parameter */
	lpp(&S390_lowcore.lpp);

	debug_sprintf_event(sfdbg, 6, "pmu_enable: es=%i cs=%i ed=%i cd=%i "
			    "interval:%lx tear=%p dear=%p\n",
			    cpuhw->lsctl.es, cpuhw->lsctl.cs, cpuhw->lsctl.ed,
			    cpuhw->lsctl.cd, cpuhw->lsctl.interval,
			    (void *) cpuhw->lsctl.tear,
			    (void *) cpuhw->lsctl.dear);
}

static void cpumsf_pmu_disable(struct pmu *pmu)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);
	struct hws_lsctl_request_block inactive;
	struct hws_qsi_info_block si;
	int err;

	if (!(cpuhw->flags & PMU_F_ENABLED))
		return;

	if (cpuhw->flags & PMU_F_ERR_MASK)
		return;

	/* Switch off sampling activation control */
	inactive = cpuhw->lsctl;
	inactive.cs = 0;
	inactive.cd = 0;

	err = lsctl(&inactive);
	if (err) {
		pr_err("Loading sampling controls failed: op=%i err=%i\n",
			2, err);
		return;
	}

	/* Save state of TEAR and DEAR register contents */
	if (!qsi(&si)) {
		/* TEAR/DEAR values are valid only if the sampling facility is
		 * enabled.  Note that cpumsf_pmu_disable() might be called even
		 * for a disabled sampling facility because cpumsf_pmu_enable()
		 * controls the enable/disable state.
		 */
		if (si.es) {
			cpuhw->lsctl.tear = si.tear;
			cpuhw->lsctl.dear = si.dear;
		}
	} else
		debug_sprintf_event(sfdbg, 3, "cpumsf_pmu_disable: "
				    "qsi() failed with err=%i\n", err);

	cpuhw->flags &= ~PMU_F_ENABLED;
}

/* perf_exclude_event() - Filter event
 * @event:	The perf event
 * @regs:	pt_regs structure
 * @sde_regs:	Sample-data-entry (sde) regs structure
 *
 * Filter perf events according to their exclude specification.
 *
 * Return non-zero if the event shall be excluded.
 */
static int perf_exclude_event(struct perf_event *event, struct pt_regs *regs,
			      struct perf_sf_sde_regs *sde_regs)
{
	if (event->attr.exclude_user && user_mode(regs))
		return 1;
	if (event->attr.exclude_kernel && !user_mode(regs))
		return 1;
	if (event->attr.exclude_guest && sde_regs->in_guest)
		return 1;
	if (event->attr.exclude_host && !sde_regs->in_guest)
		return 1;
	return 0;
}

/* perf_push_sample() - Push samples to perf
 * @event:	The perf event
 * @sample:	Hardware sample data
 *
 * Use the hardware sample data to create perf event sample.  The sample
 * is the pushed to the event subsystem and the function checks for
 * possible event overflows.  If an event overflow occurs, the PMU is
 * stopped.
 *
 * Return non-zero if an event overflow occurred.
 */
static int perf_push_sample(struct perf_event *event,
			    struct hws_basic_entry *basic)
{
	int overflow;
	struct pt_regs regs;
	struct perf_sf_sde_regs *sde_regs;
	struct perf_sample_data data;

	/* Setup perf sample */
	perf_sample_data_init(&data, 0, event->hw.last_period);

	/* Setup pt_regs to look like an CPU-measurement external interrupt
	 * using the Program Request Alert code.  The regs.int_parm_long
	 * field which is unused contains additional sample-data-entry related
	 * indicators.
	 */
	memset(&regs, 0, sizeof(regs));
	regs.int_code = 0x1407;
	regs.int_parm = CPU_MF_INT_SF_PRA;
	sde_regs = (struct perf_sf_sde_regs *) &regs.int_parm_long;

	psw_bits(regs.psw).ia	= basic->ia;
	psw_bits(regs.psw).dat	= basic->T;
	psw_bits(regs.psw).wait = basic->W;
	psw_bits(regs.psw).pstate = basic->P;
	psw_bits(regs.psw).as	= basic->AS;

	/*
	 * Use the hardware provided configuration level to decide if the
	 * sample belongs to a guest or host. If that is not available,
	 * fall back to the following heuristics:
	 * A non-zero guest program parameter always indicates a guest
	 * sample. Some early samples or samples from guests without
	 * lpp usage would be misaccounted to the host. We use the asn
	 * value as an addon heuristic to detect most of these guest samples.
	 * If the value differs from 0xffff (the host value), we assume to
	 * be a KVM guest.
	 */
	switch (basic->CL) {
	case 1: /* logical partition */
		sde_regs->in_guest = 0;
		break;
	case 2: /* virtual machine */
		sde_regs->in_guest = 1;
		break;
	default: /* old machine, use heuristics */
		if (basic->gpp || basic->prim_asn != 0xffff)
			sde_regs->in_guest = 1;
		break;
	}

	/*
	 * Store the PID value from the sample-data-entry to be
	 * processed and resolved by cpumsf_output_event_pid().
	 */
	data.tid_entry.pid = basic->hpp & LPP_PID_MASK;

	overflow = 0;
	if (perf_exclude_event(event, &regs, sde_regs))
		goto out;
	if (perf_event_overflow(event, &data, &regs)) {
		overflow = 1;
		event->pmu->stop(event, 0);
	}
	perf_event_update_userpage(event);
out:
	return overflow;
}

static void perf_event_count_update(struct perf_event *event, u64 count)
{
	local64_add(count, &event->count);
}

static void debug_sample_entry(struct hws_basic_entry *sample,
			       struct hws_trailer_entry *te)
{
	debug_sprintf_event(sfdbg, 4, "hw_collect_samples: Found unknown "
			    "sampling data entry: te->f=%i basic.def=%04x "
			    "(%p)\n",
			    te->f, sample->def, sample);
}

/* hw_collect_samples() - Walk through a sample-data-block and collect samples
 * @event:	The perf event
 * @sdbt:	Sample-data-block table
 * @overflow:	Event overflow counter
 *
 * Walks through a sample-data-block and collects sampling data entries that are
 * then pushed to the perf event subsystem.  Depending on the sampling function,
 * there can be either basic-sampling or combined-sampling data entries.  A
 * combined-sampling data entry consists of a basic- and a diagnostic-sampling
 * data entry.	The sampling function is determined by the flags in the perf
 * event hardware structure.  The function always works with a combined-sampling
 * data entry but ignores the the diagnostic portion if it is not available.
 *
 * Note that the implementation focuses on basic-sampling data entries and, if
 * such an entry is not valid, the entire combined-sampling data entry is
 * ignored.
 *
 * The overflow variables counts the number of samples that has been discarded
 * due to a perf event overflow.
 */
static void hw_collect_samples(struct perf_event *event, unsigned long *sdbt,
			       unsigned long long *overflow)
{
	struct hws_trailer_entry *te;
	struct hws_basic_entry *sample;

	te = (struct hws_trailer_entry *) trailer_entry_ptr(*sdbt);
	sample = (struct hws_basic_entry *) *sdbt;
	while ((unsigned long *) sample < (unsigned long *) te) {
		/* Check for an empty sample */
		if (!sample->def)
			break;

		/* Update perf event period */
		perf_event_count_update(event, SAMPL_RATE(&event->hw));

		/* Check whether sample is valid */
		if (sample->def == 0x0001) {
			/* If an event overflow occurred, the PMU is stopped to
			 * throttle event delivery.  Remaining sample data is
			 * discarded.
			 */
			if (!*overflow) {
				/* Check whether sample is consistent */
				if (sample->I == 0 && sample->W == 0) {
					/* Deliver sample data to perf */
					*overflow = perf_push_sample(event,
								     sample);
				}
			} else
				/* Count discarded samples */
				*overflow += 1;
		} else {
			debug_sample_entry(sample, te);
			/* Sample slot is not yet written or other record.
			 *
			 * This condition can occur if the buffer was reused
			 * from a combined basic- and diagnostic-sampling.
			 * If only basic-sampling is then active, entries are
			 * written into the larger diagnostic entries.
			 * This is typically the case for sample-data-blocks
			 * that are not full.  Stop processing if the first
			 * invalid format was detected.
			 */
			if (!te->f)
				break;
		}

		/* Reset sample slot and advance to next sample */
		sample->def = 0;
		sample++;
	}
}

/* hw_perf_event_update() - Process sampling buffer
 * @event:	The perf event
 * @flush_all:	Flag to also flush partially filled sample-data-blocks
 *
 * Processes the sampling buffer and create perf event samples.
 * The sampling buffer position are retrieved and saved in the TEAR_REG
 * register of the specified perf event.
 *
 * Only full sample-data-blocks are processed.	Specify the flash_all flag
 * to also walk through partially filled sample-data-blocks.  It is ignored
 * if PERF_CPUM_SF_FULL_BLOCKS is set.	The PERF_CPUM_SF_FULL_BLOCKS flag
 * enforces the processing of full sample-data-blocks only (trailer entries
 * with the block-full-indicator bit set).
 */
static void hw_perf_event_update(struct perf_event *event, int flush_all)
{
	struct hw_perf_event *hwc = &event->hw;
	struct hws_trailer_entry *te;
	unsigned long *sdbt;
	unsigned long long event_overflow, sampl_overflow, num_sdb, te_flags;
	int done;

	/*
	 * AUX buffer is used when in diagnostic sampling mode.
	 * No perf events/samples are created.
	 */
	if (SAMPL_DIAG_MODE(&event->hw))
		return;

	if (flush_all && SDB_FULL_BLOCKS(hwc))
		flush_all = 0;

	sdbt = (unsigned long *) TEAR_REG(hwc);
	done = event_overflow = sampl_overflow = num_sdb = 0;
	while (!done) {
		/* Get the trailer entry of the sample-data-block */
		te = (struct hws_trailer_entry *) trailer_entry_ptr(*sdbt);

		/* Leave loop if no more work to do (block full indicator) */
		if (!te->f) {
			done = 1;
			if (!flush_all)
				break;
		}

		/* Check the sample overflow count */
		if (te->overflow)
			/* Account sample overflows and, if a particular limit
			 * is reached, extend the sampling buffer.
			 * For details, see sfb_account_overflows().
			 */
			sampl_overflow += te->overflow;

		/* Timestamps are valid for full sample-data-blocks only */
		debug_sprintf_event(sfdbg, 6, "hw_perf_event_update: sdbt=%p "
				    "overflow=%llu timestamp=%#llx\n",
				    sdbt, te->overflow,
				    (te->f) ? trailer_timestamp(te) : 0ULL);

		/* Collect all samples from a single sample-data-block and
		 * flag if an (perf) event overflow happened.  If so, the PMU
		 * is stopped and remaining samples will be discarded.
		 */
		hw_collect_samples(event, sdbt, &event_overflow);
		num_sdb++;

		/* Reset trailer (using compare-double-and-swap) */
		do {
			te_flags = te->flags & ~SDB_TE_BUFFER_FULL_MASK;
			te_flags |= SDB_TE_ALERT_REQ_MASK;
		} while (!cmpxchg_double(&te->flags, &te->overflow,
					 te->flags, te->overflow,
					 te_flags, 0ULL));

		/* Advance to next sample-data-block */
		sdbt++;
		if (is_link_entry(sdbt))
			sdbt = get_next_sdbt(sdbt);

		/* Update event hardware registers */
		TEAR_REG(hwc) = (unsigned long) sdbt;

		/* Stop processing sample-data if all samples of the current
		 * sample-data-block were flushed even if it was not full.
		 */
		if (flush_all && done)
			break;

		/* If an event overflow happened, discard samples by
		 * processing any remaining sample-data-blocks.
		 */
		if (event_overflow)
			flush_all = 1;
	}

	/* Account sample overflows in the event hardware structure */
	if (sampl_overflow)
		OVERFLOW_REG(hwc) = DIV_ROUND_UP(OVERFLOW_REG(hwc) +
						 sampl_overflow, 1 + num_sdb);
	if (sampl_overflow || event_overflow)
		debug_sprintf_event(sfdbg, 4, "hw_perf_event_update: "
				    "overflow stats: sample=%llu event=%llu\n",
				    sampl_overflow, event_overflow);
}

#define AUX_SDB_INDEX(aux, i) ((i) % aux->sfb.num_sdb)
#define AUX_SDB_NUM(aux, start, end) (end >= start ? end - start + 1 : 0)
#define AUX_SDB_NUM_ALERT(aux) AUX_SDB_NUM(aux, aux->head, aux->alert_mark)
#define AUX_SDB_NUM_EMPTY(aux) AUX_SDB_NUM(aux, aux->head, aux->empty_mark)

/*
 * Get trailer entry by index of SDB.
 */
static struct hws_trailer_entry *aux_sdb_trailer(struct aux_buffer *aux,
						 unsigned long index)
{
	unsigned long sdb;

	index = AUX_SDB_INDEX(aux, index);
	sdb = aux->sdb_index[index];
	return (struct hws_trailer_entry *)trailer_entry_ptr(sdb);
}

/*
 * Finish sampling on the cpu. Called by cpumsf_pmu_del() with pmu
 * disabled. Collect the full SDBs in AUX buffer which have not reached
 * the point of alert indicator. And ignore the SDBs which are not
 * full.
 *
 * 1. Scan SDBs to see how much data is there and consume them.
 * 2. Remove alert indicator in the buffer.
 */
static void aux_output_end(struct perf_output_handle *handle)
{
	unsigned long i, range_scan, idx;
	struct aux_buffer *aux;
	struct hws_trailer_entry *te;

	aux = perf_get_aux(handle);
	if (!aux)
		return;

	range_scan = AUX_SDB_NUM_ALERT(aux);
	for (i = 0, idx = aux->head; i < range_scan; i++, idx++) {
		te = aux_sdb_trailer(aux, idx);
		if (!(te->flags & SDB_TE_BUFFER_FULL_MASK))
			break;
	}
	/* i is num of SDBs which are full */
	perf_aux_output_end(handle, i << PAGE_SHIFT);

	/* Remove alert indicators in the buffer */
	te = aux_sdb_trailer(aux, aux->alert_mark);
	te->flags &= ~SDB_TE_ALERT_REQ_MASK;

	debug_sprintf_event(sfdbg, 6, "aux_output_end: collect %lx SDBs\n", i);
}

/*
 * Start sampling on the CPU. Called by cpumsf_pmu_add() when an event
 * is first added to the CPU or rescheduled again to the CPU. It is called
 * with pmu disabled.
 *
 * 1. Reset the trailer of SDBs to get ready for new data.
 * 2. Tell the hardware where to put the data by reset the SDBs buffer
 *    head(tear/dear).
 */
static int aux_output_begin(struct perf_output_handle *handle,
			    struct aux_buffer *aux,
			    struct cpu_hw_sf *cpuhw)
{
	unsigned long range;
	unsigned long i, range_scan, idx;
	unsigned long head, base, offset;
	struct hws_trailer_entry *te;

	if (WARN_ON_ONCE(handle->head & ~PAGE_MASK))
		return -EINVAL;

	aux->head = handle->head >> PAGE_SHIFT;
	range = (handle->size + 1) >> PAGE_SHIFT;
	if (range <= 1)
		return -ENOMEM;

	/*
	 * SDBs between aux->head and aux->empty_mark are already ready
	 * for new data. range_scan is num of SDBs not within them.
	 */
	if (range > AUX_SDB_NUM_EMPTY(aux)) {
		range_scan = range - AUX_SDB_NUM_EMPTY(aux);
		idx = aux->empty_mark + 1;
		for (i = 0; i < range_scan; i++, idx++) {
			te = aux_sdb_trailer(aux, idx);
			te->flags = te->flags & ~SDB_TE_BUFFER_FULL_MASK;
			te->flags = te->flags & ~SDB_TE_ALERT_REQ_MASK;
			te->overflow = 0;
		}
		/* Save the position of empty SDBs */
		aux->empty_mark = aux->head + range - 1;
	}

	/* Set alert indicator */
	aux->alert_mark = aux->head + range/2 - 1;
	te = aux_sdb_trailer(aux, aux->alert_mark);
	te->flags = te->flags | SDB_TE_ALERT_REQ_MASK;

	/* Reset hardware buffer head */
	head = AUX_SDB_INDEX(aux, aux->head);
	base = aux->sdbt_index[head / CPUM_SF_SDB_PER_TABLE];
	offset = head % CPUM_SF_SDB_PER_TABLE;
	cpuhw->lsctl.tear = base + offset * sizeof(unsigned long);
	cpuhw->lsctl.dear = aux->sdb_index[head];

	debug_sprintf_event(sfdbg, 6, "aux_output_begin: "
			    "head->alert_mark->empty_mark (num_alert, range)"
			    "[%lx -> %lx -> %lx] (%lx, %lx) "
			    "tear index %lx, tear %lx dear %lx\n",
			    aux->head, aux->alert_mark, aux->empty_mark,
			    AUX_SDB_NUM_ALERT(aux), range,
			    head / CPUM_SF_SDB_PER_TABLE,
			    cpuhw->lsctl.tear,
			    cpuhw->lsctl.dear);

	return 0;
}

/*
 * Set alert indicator on SDB at index @alert_index while sampler is running.
 *
 * Return true if successfully.
 * Return false if full indicator is already set by hardware sampler.
 */
static bool aux_set_alert(struct aux_buffer *aux, unsigned long alert_index,
			  unsigned long long *overflow)
{
	unsigned long long orig_overflow, orig_flags, new_flags;
	struct hws_trailer_entry *te;

	te = aux_sdb_trailer(aux, alert_index);
	do {
		orig_flags = te->flags;
		orig_overflow = te->overflow;
		*overflow = orig_overflow;
		if (orig_flags & SDB_TE_BUFFER_FULL_MASK) {
			/*
			 * SDB is already set by hardware.
			 * Abort and try to set somewhere
			 * behind.
			 */
			return false;
		}
		new_flags = orig_flags | SDB_TE_ALERT_REQ_MASK;
	} while (!cmpxchg_double(&te->flags, &te->overflow,
				 orig_flags, orig_overflow,
				 new_flags, 0ULL));
	return true;
}

/*
 * aux_reset_buffer() - Scan and setup SDBs for new samples
 * @aux:	The AUX buffer to set
 * @range:	The range of SDBs to scan started from aux->head
 * @overflow:	Set to overflow count
 *
 * Set alert indicator on the SDB at index of aux->alert_mark. If this SDB is
 * marked as empty, check if it is already set full by the hardware sampler.
 * If yes, that means new data is already there before we can set an alert
 * indicator. Caller should try to set alert indicator to some position behind.
 *
 * Scan the SDBs in AUX buffer from behind aux->empty_mark. They are used
 * previously and have already been consumed by user space. Reset these SDBs
 * (clear full indicator and alert indicator) for new data.
 * If aux->alert_mark fall in this area, just set it. Overflow count is
 * recorded while scanning.
 *
 * SDBs between aux->head and aux->empty_mark are already reset at last time.
 * and ready for new samples. So scanning on this area could be skipped.
 *
 * Return true if alert indicator is set successfully and false if not.
 */
static bool aux_reset_buffer(struct aux_buffer *aux, unsigned long range,
			     unsigned long long *overflow)
{
	unsigned long long orig_overflow, orig_flags, new_flags;
	unsigned long i, range_scan, idx;
	struct hws_trailer_entry *te;

	if (range <= AUX_SDB_NUM_EMPTY(aux))
		/*
		 * No need to scan. All SDBs in range are marked as empty.
		 * Just set alert indicator. Should check race with hardware
		 * sampler.
		 */
		return aux_set_alert(aux, aux->alert_mark, overflow);

	if (aux->alert_mark <= aux->empty_mark)
		/*
		 * Set alert indicator on empty SDB. Should check race
		 * with hardware sampler.
		 */
		if (!aux_set_alert(aux, aux->alert_mark, overflow))
			return false;

	/*
	 * Scan the SDBs to clear full and alert indicator used previously.
	 * Start scanning from one SDB behind empty_mark. If the new alert
	 * indicator fall into this range, set it.
	 */
	range_scan = range - AUX_SDB_NUM_EMPTY(aux);
	idx = aux->empty_mark + 1;
	for (i = 0; i < range_scan; i++, idx++) {
		te = aux_sdb_trailer(aux, idx);
		do {
			orig_flags = te->flags;
			orig_overflow = te->overflow;
			new_flags = orig_flags & ~SDB_TE_BUFFER_FULL_MASK;
			if (idx == aux->alert_mark)
				new_flags |= SDB_TE_ALERT_REQ_MASK;
			else
				new_flags &= ~SDB_TE_ALERT_REQ_MASK;
		} while (!cmpxchg_double(&te->flags, &te->overflow,
					 orig_flags, orig_overflow,
					 new_flags, 0ULL));
		*overflow += orig_overflow;
	}

	/* Update empty_mark to new position */
	aux->empty_mark = aux->head + range - 1;

	return true;
}

/*
 * Measurement alert handler for diagnostic mode sampling.
 */
static void hw_collect_aux(struct cpu_hw_sf *cpuhw)
{
	struct aux_buffer *aux;
	int done = 0;
	unsigned long range = 0, size;
	unsigned long long overflow = 0;
	struct perf_output_handle *handle = &cpuhw->handle;
	unsigned long num_sdb;

	aux = perf_get_aux(handle);
	if (WARN_ON_ONCE(!aux))
		return;

	/* Inform user space new data arrived */
	size = AUX_SDB_NUM_ALERT(aux) << PAGE_SHIFT;
	perf_aux_output_end(handle, size);
	num_sdb = aux->sfb.num_sdb;

	while (!done) {
		/* Get an output handle */
		aux = perf_aux_output_begin(handle, cpuhw->event);
		if (handle->size == 0) {
			pr_err("The AUX buffer with %lu pages for the "
			       "diagnostic-sampling mode is full\n",
				num_sdb);
			debug_sprintf_event(sfdbg, 1, "AUX buffer used up\n");
			break;
		}
		if (WARN_ON_ONCE(!aux))
			return;

		/* Update head and alert_mark to new position */
		aux->head = handle->head >> PAGE_SHIFT;
		range = (handle->size + 1) >> PAGE_SHIFT;
		if (range == 1)
			aux->alert_mark = aux->head;
		else
			aux->alert_mark = aux->head + range/2 - 1;

		if (aux_reset_buffer(aux, range, &overflow)) {
			if (!overflow) {
				done = 1;
				break;
			}
			size = range << PAGE_SHIFT;
			perf_aux_output_end(&cpuhw->handle, size);
			pr_err("Sample data caused the AUX buffer with %lu "
			       "pages to overflow\n", num_sdb);
			debug_sprintf_event(sfdbg, 1, "head %lx range %lx "
					    "overflow %llx\n",
					    aux->head, range, overflow);
		} else {
			size = AUX_SDB_NUM_ALERT(aux) << PAGE_SHIFT;
			perf_aux_output_end(&cpuhw->handle, size);
			debug_sprintf_event(sfdbg, 6, "head %lx alert %lx "
					    "already full, try another\n",
					    aux->head, aux->alert_mark);
		}
	}

	if (done)
		debug_sprintf_event(sfdbg, 6, "aux_reset_buffer: "
				    "[%lx -> %lx -> %lx] (%lx, %lx)\n",
				    aux->head, aux->alert_mark, aux->empty_mark,
				    AUX_SDB_NUM_ALERT(aux), range);
}

/*
 * Callback when freeing AUX buffers.
 */
static void aux_buffer_free(void *data)
{
	struct aux_buffer *aux = data;
	unsigned long i, num_sdbt;

	if (!aux)
		return;

	/* Free SDBT. SDB is freed by the caller */
	num_sdbt = aux->sfb.num_sdbt;
	for (i = 0; i < num_sdbt; i++)
		free_page(aux->sdbt_index[i]);

	kfree(aux->sdbt_index);
	kfree(aux->sdb_index);
	kfree(aux);

	debug_sprintf_event(sfdbg, 4, "aux_buffer_free: free "
			    "%lu SDBTs\n", num_sdbt);
}

static void aux_sdb_init(unsigned long sdb)
{
	struct hws_trailer_entry *te;

	te = (struct hws_trailer_entry *)trailer_entry_ptr(sdb);

	/* Save clock base */
	te->clock_base = 1;
	memcpy(&te->progusage2, &tod_clock_base[1], 8);
}

/*
 * aux_buffer_setup() - Setup AUX buffer for diagnostic mode sampling
 * @event:	Event the buffer is setup for, event->cpu == -1 means current
 * @pages:	Array of pointers to buffer pages passed from perf core
 * @nr_pages:	Total pages
 * @snapshot:	Flag for snapshot mode
 *
 * This is the callback when setup an event using AUX buffer. Perf tool can
 * trigger this by an additional mmap() call on the event. Unlike the buffer
 * for basic samples, AUX buffer belongs to the event. It is scheduled with
 * the task among online cpus when it is a per-thread event.
 *
 * Return the private AUX buffer structure if success or NULL if fails.
 */
static void *aux_buffer_setup(struct perf_event *event, void **pages,
			      int nr_pages, bool snapshot)
{
	struct sf_buffer *sfb;
	struct aux_buffer *aux;
	unsigned long *new, *tail;
	int i, n_sdbt;

	if (!nr_pages || !pages)
		return NULL;

	if (nr_pages > CPUM_SF_MAX_SDB * CPUM_SF_SDB_DIAG_FACTOR) {
		pr_err("AUX buffer size (%i pages) is larger than the "
		       "maximum sampling buffer limit\n",
		       nr_pages);
		return NULL;
	} else if (nr_pages < CPUM_SF_MIN_SDB * CPUM_SF_SDB_DIAG_FACTOR) {
		pr_err("AUX buffer size (%i pages) is less than the "
		       "minimum sampling buffer limit\n",
		       nr_pages);
		return NULL;
	}

	/* Allocate aux_buffer struct for the event */
	aux = kmalloc(sizeof(struct aux_buffer), GFP_KERNEL);
	if (!aux)
		goto no_aux;
	sfb = &aux->sfb;

	/* Allocate sdbt_index for fast reference */
	n_sdbt = (nr_pages + CPUM_SF_SDB_PER_TABLE - 1) / CPUM_SF_SDB_PER_TABLE;
	aux->sdbt_index = kmalloc_array(n_sdbt, sizeof(void *), GFP_KERNEL);
	if (!aux->sdbt_index)
		goto no_sdbt_index;

	/* Allocate sdb_index for fast reference */
	aux->sdb_index = kmalloc_array(nr_pages, sizeof(void *), GFP_KERNEL);
	if (!aux->sdb_index)
		goto no_sdb_index;

	/* Allocate the first SDBT */
	sfb->num_sdbt = 0;
	sfb->sdbt = (unsigned long *) get_zeroed_page(GFP_KERNEL);
	if (!sfb->sdbt)
		goto no_sdbt;
	aux->sdbt_index[sfb->num_sdbt++] = (unsigned long)sfb->sdbt;
	tail = sfb->tail = sfb->sdbt;

	/*
	 * Link the provided pages of AUX buffer to SDBT.
	 * Allocate SDBT if needed.
	 */
	for (i = 0; i < nr_pages; i++, tail++) {
		if (require_table_link(tail)) {
			new = (unsigned long *) get_zeroed_page(GFP_KERNEL);
			if (!new)
				goto no_sdbt;
			aux->sdbt_index[sfb->num_sdbt++] = (unsigned long)new;
			/* Link current page to tail of chain */
			*tail = (unsigned long)(void *) new + 1;
			tail = new;
		}
		/* Tail is the entry in a SDBT */
		*tail = (unsigned long)pages[i];
		aux->sdb_index[i] = (unsigned long)pages[i];
		aux_sdb_init((unsigned long)pages[i]);
	}
	sfb->num_sdb = nr_pages;

	/* Link the last entry in the SDBT to the first SDBT */
	*tail = (unsigned long) sfb->sdbt + 1;
	sfb->tail = tail;

	/*
	 * Initial all SDBs are zeroed. Mark it as empty.
	 * So there is no need to clear the full indicator
	 * when this event is first added.
	 */
	aux->empty_mark = sfb->num_sdb - 1;

	debug_sprintf_event(sfdbg, 4, "aux_buffer_setup: setup %lu SDBTs"
			    " and %lu SDBs\n",
			    sfb->num_sdbt, sfb->num_sdb);

	return aux;

no_sdbt:
	/* SDBs (AUX buffer pages) are freed by caller */
	for (i = 0; i < sfb->num_sdbt; i++)
		free_page(aux->sdbt_index[i]);
	kfree(aux->sdb_index);
no_sdb_index:
	kfree(aux->sdbt_index);
no_sdbt_index:
	kfree(aux);
no_aux:
	return NULL;
}

static void cpumsf_pmu_read(struct perf_event *event)
{
	/* Nothing to do ... updates are interrupt-driven */
}

/* Check if the new sampling period/freqeuncy is appropriate.
 *
 * Return non-zero on error and zero on passed checks.
 */
static int cpumsf_pmu_check_period(struct perf_event *event, u64 value)
{
	struct hws_qsi_info_block si;
	unsigned long rate;
	bool do_freq;

	memset(&si, 0, sizeof(si));
	if (event->cpu == -1) {
		if (qsi(&si))
			return -ENODEV;
	} else {
		/* Event is pinned to a particular CPU, retrieve the per-CPU
		 * sampling structure for accessing the CPU-specific QSI.
		 */
		struct cpu_hw_sf *cpuhw = &per_cpu(cpu_hw_sf, event->cpu);

		si = cpuhw->qsi;
	}

	do_freq = !!SAMPLE_FREQ_MODE(&event->hw);
	rate = getrate(do_freq, value, &si);
	if (!rate)
		return -EINVAL;

	event->attr.sample_period = rate;
	SAMPL_RATE(&event->hw) = rate;
	hw_init_period(&event->hw, SAMPL_RATE(&event->hw));
	debug_sprintf_event(sfdbg, 4, "cpumsf_pmu_check_period:"
			    "cpu:%d value:%llx period:%llx freq:%d\n",
			    event->cpu, value,
			    event->attr.sample_period, do_freq);
	return 0;
}

/* Activate sampling control.
 * Next call of pmu_enable() starts sampling.
 */
static void cpumsf_pmu_start(struct perf_event *event, int flags)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);

	if (WARN_ON_ONCE(!(event->hw.state & PERF_HES_STOPPED)))
		return;

	if (flags & PERF_EF_RELOAD)
		WARN_ON_ONCE(!(event->hw.state & PERF_HES_UPTODATE));

	perf_pmu_disable(event->pmu);
	event->hw.state = 0;
	cpuhw->lsctl.cs = 1;
	if (SAMPL_DIAG_MODE(&event->hw))
		cpuhw->lsctl.cd = 1;
	perf_pmu_enable(event->pmu);
}

/* Deactivate sampling control.
 * Next call of pmu_enable() stops sampling.
 */
static void cpumsf_pmu_stop(struct perf_event *event, int flags)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);

	if (event->hw.state & PERF_HES_STOPPED)
		return;

	perf_pmu_disable(event->pmu);
	cpuhw->lsctl.cs = 0;
	cpuhw->lsctl.cd = 0;
	event->hw.state |= PERF_HES_STOPPED;

	if ((flags & PERF_EF_UPDATE) && !(event->hw.state & PERF_HES_UPTODATE)) {
		hw_perf_event_update(event, 1);
		event->hw.state |= PERF_HES_UPTODATE;
	}
	perf_pmu_enable(event->pmu);
}

static int cpumsf_pmu_add(struct perf_event *event, int flags)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);
	struct aux_buffer *aux;
	int err;

	if (cpuhw->flags & PMU_F_IN_USE)
		return -EAGAIN;

	if (!SAMPL_DIAG_MODE(&event->hw) && !cpuhw->sfb.sdbt)
		return -EINVAL;

	err = 0;
	perf_pmu_disable(event->pmu);

	event->hw.state = PERF_HES_UPTODATE | PERF_HES_STOPPED;

	/* Set up sampling controls.  Always program the sampling register
	 * using the SDB-table start.  Reset TEAR_REG event hardware register
	 * that is used by hw_perf_event_update() to store the sampling buffer
	 * position after samples have been flushed.
	 */
	cpuhw->lsctl.s = 0;
	cpuhw->lsctl.h = 1;
	cpuhw->lsctl.interval = SAMPL_RATE(&event->hw);
	if (!SAMPL_DIAG_MODE(&event->hw)) {
		cpuhw->lsctl.tear = (unsigned long) cpuhw->sfb.sdbt;
		cpuhw->lsctl.dear = *(unsigned long *) cpuhw->sfb.sdbt;
		hw_reset_registers(&event->hw, cpuhw->sfb.sdbt);
	}

	/* Ensure sampling functions are in the disabled state.  If disabled,
	 * switch on sampling enable control. */
	if (WARN_ON_ONCE(cpuhw->lsctl.es == 1 || cpuhw->lsctl.ed == 1)) {
		err = -EAGAIN;
		goto out;
	}
	if (SAMPL_DIAG_MODE(&event->hw)) {
		aux = perf_aux_output_begin(&cpuhw->handle, event);
		if (!aux) {
			err = -EINVAL;
			goto out;
		}
		err = aux_output_begin(&cpuhw->handle, aux, cpuhw);
		if (err)
			goto out;
		cpuhw->lsctl.ed = 1;
	}
	cpuhw->lsctl.es = 1;

	/* Set in_use flag and store event */
	cpuhw->event = event;
	cpuhw->flags |= PMU_F_IN_USE;

	if (flags & PERF_EF_START)
		cpumsf_pmu_start(event, PERF_EF_RELOAD);
out:
	perf_event_update_userpage(event);
	perf_pmu_enable(event->pmu);
	return err;
}

static void cpumsf_pmu_del(struct perf_event *event, int flags)
{
	struct cpu_hw_sf *cpuhw = this_cpu_ptr(&cpu_hw_sf);

	perf_pmu_disable(event->pmu);
	cpumsf_pmu_stop(event, PERF_EF_UPDATE);

	cpuhw->lsctl.es = 0;
	cpuhw->lsctl.ed = 0;
	cpuhw->flags &= ~PMU_F_IN_USE;
	cpuhw->event = NULL;

	if (SAMPL_DIAG_MODE(&event->hw))
		aux_output_end(&cpuhw->handle);
	perf_event_update_userpage(event);
	perf_pmu_enable(event->pmu);
}

CPUMF_EVENT_ATTR(SF, SF_CYCLES_BASIC, PERF_EVENT_CPUM_SF);
CPUMF_EVENT_ATTR(SF, SF_CYCLES_BASIC_DIAG, PERF_EVENT_CPUM_SF_DIAG);

/* Attribute list for CPU_SF.
 *
 * The availablitiy depends on the CPU_MF sampling facility authorization
 * for basic + diagnositic samples. This is determined at initialization
 * time by the sampling facility device driver.
 * If the authorization for basic samples is turned off, it should be
 * also turned off for diagnostic sampling.
 *
 * During initialization of the device driver, check the authorization
 * level for diagnostic sampling and installs the attribute
 * file for diagnostic sampling if necessary.
 *
 * For now install a placeholder to reference all possible attributes:
 * SF_CYCLES_BASIC and SF_CYCLES_BASIC_DIAG.
 * Add another entry for the final NULL pointer.
 */
enum {
	SF_CYCLES_BASIC_ATTR_IDX = 0,
	SF_CYCLES_BASIC_DIAG_ATTR_IDX,
	SF_CYCLES_ATTR_MAX
};

static struct attribute *cpumsf_pmu_events_attr[SF_CYCLES_ATTR_MAX + 1] = {
	[SF_CYCLES_BASIC_ATTR_IDX] = CPUMF_EVENT_PTR(SF, SF_CYCLES_BASIC)
};

PMU_FORMAT_ATTR(event, "config:0-63");

static struct attribute *cpumsf_pmu_format_attr[] = {
	&format_attr_event.attr,
	NULL,
};

static struct attribute_group cpumsf_pmu_events_group = {
	.name = "events",
	.attrs = cpumsf_pmu_events_attr,
};

static struct attribute_group cpumsf_pmu_format_group = {
	.name = "format",
	.attrs = cpumsf_pmu_format_attr,
};

static const struct attribute_group *cpumsf_pmu_attr_groups[] = {
	&cpumsf_pmu_events_group,
	&cpumsf_pmu_format_group,
	NULL,
};

static struct pmu cpumf_sampling = {
	.pmu_enable   = cpumsf_pmu_enable,
	.pmu_disable  = cpumsf_pmu_disable,

	.event_init   = cpumsf_pmu_event_init,
	.add	      = cpumsf_pmu_add,
	.del	      = cpumsf_pmu_del,

	.start	      = cpumsf_pmu_start,
	.stop	      = cpumsf_pmu_stop,
	.read	      = cpumsf_pmu_read,

	.attr_groups  = cpumsf_pmu_attr_groups,

	.setup_aux    = aux_buffer_setup,
	.free_aux     = aux_buffer_free,

	.check_period = cpumsf_pmu_check_period,
};

static void cpumf_measurement_alert(struct ext_code ext_code,
				    unsigned int alert, unsigned long unused)
{
	struct cpu_hw_sf *cpuhw;

	if (!(alert & CPU_MF_INT_SF_MASK))
		return;
	inc_irq_stat(IRQEXT_CMS);
	cpuhw = this_cpu_ptr(&cpu_hw_sf);

	/* Measurement alerts are shared and might happen when the PMU
	 * is not reserved.  Ignore these alerts in this case. */
	if (!(cpuhw->flags & PMU_F_RESERVED))
		return;

	/* The processing below must take care of multiple alert events that
	 * might be indicated concurrently. */

	/* Program alert request */
	if (alert & CPU_MF_INT_SF_PRA) {
		if (cpuhw->flags & PMU_F_IN_USE)
			if (SAMPL_DIAG_MODE(&cpuhw->event->hw))
				hw_collect_aux(cpuhw);
			else
				hw_perf_event_update(cpuhw->event, 0);
		else
			WARN_ON_ONCE(!(cpuhw->flags & PMU_F_IN_USE));
	}

	/* Report measurement alerts only for non-PRA codes */
	if (alert != CPU_MF_INT_SF_PRA)
		debug_sprintf_event(sfdbg, 6, "measurement alert: %#x\n",
				    alert);

	/* Sampling authorization change request */
	if (alert & CPU_MF_INT_SF_SACA)
		qsi(&cpuhw->qsi);

	/* Loss of sample data due to high-priority machine activities */
	if (alert & CPU_MF_INT_SF_LSDA) {
		pr_err("Sample data was lost\n");
		cpuhw->flags |= PMU_F_ERR_LSDA;
		sf_disable();
	}

	/* Invalid sampling buffer entry */
	if (alert & (CPU_MF_INT_SF_IAE|CPU_MF_INT_SF_ISE)) {
		pr_err("A sampling buffer entry is incorrect (alert=0x%x)\n",
		       alert);
		cpuhw->flags |= PMU_F_ERR_IBE;
		sf_disable();
	}
}

static int cpusf_pmu_setup(unsigned int cpu, int flags)
{
	/* Ignore the notification if no events are scheduled on the PMU.
	 * This might be racy...
	 */
	if (!atomic_read(&num_events))
		return 0;

	local_irq_disable();
	setup_pmc_cpu(&flags);
	local_irq_enable();
	return 0;
}

static int s390_pmu_sf_online_cpu(unsigned int cpu)
{
	return cpusf_pmu_setup(cpu, PMC_INIT);
}

static int s390_pmu_sf_offline_cpu(unsigned int cpu)
{
	return cpusf_pmu_setup(cpu, PMC_RELEASE);
}

static int param_get_sfb_size(char *buffer, const struct kernel_param *kp)
{
	if (!cpum_sf_avail())
		return -ENODEV;
	return sprintf(buffer, "%lu,%lu", CPUM_SF_MIN_SDB, CPUM_SF_MAX_SDB);
}

static int param_set_sfb_size(const char *val, const struct kernel_param *kp)
{
	int rc;
	unsigned long min, max;

	if (!cpum_sf_avail())
		return -ENODEV;
	if (!val || !strlen(val))
		return -EINVAL;

	/* Valid parameter values: "min,max" or "max" */
	min = CPUM_SF_MIN_SDB;
	max = CPUM_SF_MAX_SDB;
	if (strchr(val, ','))
		rc = (sscanf(val, "%lu,%lu", &min, &max) == 2) ? 0 : -EINVAL;
	else
		rc = kstrtoul(val, 10, &max);

	if (min < 2 || min >= max || max > get_num_physpages())
		rc = -EINVAL;
	if (rc)
		return rc;

	sfb_set_limits(min, max);
	pr_info("The sampling buffer limits have changed to: "
		"min=%lu max=%lu (diag=x%lu)\n",
		CPUM_SF_MIN_SDB, CPUM_SF_MAX_SDB, CPUM_SF_SDB_DIAG_FACTOR);
	return 0;
}

#define param_check_sfb_size(name, p) __param_check(name, p, void)
static const struct kernel_param_ops param_ops_sfb_size = {
	.set = param_set_sfb_size,
	.get = param_get_sfb_size,
};

#define RS_INIT_FAILURE_QSI	  0x0001
#define RS_INIT_FAILURE_BSDES	  0x0002
#define RS_INIT_FAILURE_ALRT	  0x0003
#define RS_INIT_FAILURE_PERF	  0x0004
static void __init pr_cpumsf_err(unsigned int reason)
{
	pr_err("Sampling facility support for perf is not available: "
	       "reason=%04x\n", reason);
}

static int __init init_cpum_sampling_pmu(void)
{
	struct hws_qsi_info_block si;
	int err;

	if (!cpum_sf_avail())
		return -ENODEV;

	memset(&si, 0, sizeof(si));
	if (qsi(&si)) {
		pr_cpumsf_err(RS_INIT_FAILURE_QSI);
		return -ENODEV;
	}

	if (!si.as && !si.ad)
		return -ENODEV;

	if (si.bsdes != sizeof(struct hws_basic_entry)) {
		pr_cpumsf_err(RS_INIT_FAILURE_BSDES);
		return -EINVAL;
	}

	if (si.ad) {
		sfb_set_limits(CPUM_SF_MIN_SDB, CPUM_SF_MAX_SDB);
		/* Sampling of diagnostic data authorized,
		 * install event into attribute list of PMU device.
		 */
		cpumsf_pmu_events_attr[SF_CYCLES_BASIC_DIAG_ATTR_IDX] =
			CPUMF_EVENT_PTR(SF, SF_CYCLES_BASIC_DIAG);
	}

	sfdbg = debug_register(KMSG_COMPONENT, 2, 1, 80);
	if (!sfdbg) {
		pr_err("Registering for s390dbf failed\n");
		return -ENOMEM;
	}
	debug_register_view(sfdbg, &debug_sprintf_view);

	err = register_external_irq(EXT_IRQ_MEASURE_ALERT,
				    cpumf_measurement_alert);
	if (err) {
		pr_cpumsf_err(RS_INIT_FAILURE_ALRT);
		debug_unregister(sfdbg);
		goto out;
	}

	err = perf_pmu_register(&cpumf_sampling, "cpum_sf", PERF_TYPE_RAW);
	if (err) {
		pr_cpumsf_err(RS_INIT_FAILURE_PERF);
		unregister_external_irq(EXT_IRQ_MEASURE_ALERT,
					cpumf_measurement_alert);
		debug_unregister(sfdbg);
		goto out;
	}

	cpuhp_setup_state(CPUHP_AP_PERF_S390_SF_ONLINE, "perf/s390/sf:online",
			  s390_pmu_sf_online_cpu, s390_pmu_sf_offline_cpu);
out:
	return err;
}

arch_initcall(init_cpum_sampling_pmu);
core_param(cpum_sfb_size, CPUM_SF_MAX_SDB, sfb_size, 0640);